SEO & marketing · checked 2026-10-19

Can I vibecode LLM Pulse?

KINDA· weekend project

Asking a few AI models your prompts on a schedule, saving the answers and counting brand mentions and cited sites is a weekend project on SQLite with the OpenAI, Perplexity and Gemini APIs. The catch is what you measure: LLM Pulse's own blog says it runs prompts through the real chat interfaces and that tools built on APIs miss what users actually see. Google AI Overviews and AI Mode have no official API at all, so a home-built tracker is a directional proxy, not the same measurement.

$55/moprice

$660you'd save a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyIntermediate · Some system setup or keys.

TimeA weekend; prompt design and detection take the time

Claude Code, Codex or CursorAny of them can run the prompt.
Node.js 22better-sqlite3 compiles a native module, so build tools must be installed.
AI provider API keysOpenAI, Perplexity and Gemini each bill per call, and web-search calls cost more than plain ones.
A prompt list20 to 50 questions your customers would really ask, plus your competitors' names.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a local AI-visibility tracker to replace LLM Pulse. Requirements:

- Node.js 22 + TypeScript: a Fastify server and a small React dashboard built with Vite. `npm start`
serves http://localhost:4890; `npm run track` runs one tracking pass from the command line.
- Config: ./tracker.yml (`yaml`) holds your brand with aliases and domain, up to 10 competitors the same
way, and a list of prompts, each with a tag such as category, comparison or reputation.
- Providers behind one `AskProvider` interface: the OpenAI Responses API with the web_search tool,
Perplexity Sonar, and Gemini with Google Search grounding. Keys in .env, each one optional. Store the
full answer text and the cited URLs for every call.
- Each prompt runs 3 times per provider per pass because answers vary. Report mention rate as runs with a
mention divided by runs, never as a single yes or no.
- Detection: case-insensitive, word-boundary alias matching for mentions; rank is the order of first
mention among tracked brands; cited domains are counted per brand with your own domain highlighted.
- Storage: better-sqlite3 in ./data/pulse.db: prompts, runs, answers, mentions, citations. Keep every raw
answer so detection can be re-run after you fix an alias. Weekly passes with `node-cron`.
- Dashboard: mention rate and share of voice per provider over time, a prompt table with the latest
answers, top cited domains, and CSV export of every table.
- Cost guard: estimate the number of calls before a pass and refuse above MAX_PASS_COST_USD in .env.
- Optional sentiment: when a chat key is set, classify how each answer frames your brand (positive,
neutral, negative) with a fixed rubric kept in the repo.
- Single user. No accounts, no telemetry. Outbound calls: the provider APIs only.
- Out of scope: Google AI Overviews and AI Mode (no official API), scraping the chat apps, content
writing, several projects.
- README: say that API answers differ from what people see in the apps, so the numbers are directional.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own LLM Pulse.

Add a weekly email report to the AI-visibility tracker in this folder. Requirements:

- After each weekly pass, send an HTML email through nodemailer (SMTP in .env) to REPORT_TO.
- Include mention rate and share of voice per provider with the change since last week, the 5 prompts
  where you gained or lost most, and newly cited domains.
- Only flag a change when it exceeds the run-to-run variation observed in the last 4 passes; say "within noise" otherwise.
- Keep a copy as Markdown in ./reports/YYYY-MM-DD.md.
- Update the README.

Add prompt suggestions to the AI-visibility tracker in this folder. Requirements:

- A command `npm run suggest -- queries.csv` that reads a Google Search Console export (query, clicks, impressions).
- Ask the chat model to turn the top 100 queries into natural-language questions, group them by intent
  and drop duplicates of existing prompts.
- Print the candidates and add the ones I confirm to tracker.yml with a tag; never write without confirmation.
- Update the README with the Search Console export steps.

Add a citation gap view to the AI-visibility tracker in this folder. Requirements:

- A page listing domains and URLs cited in answers that mention a competitor but never mention you.
- Group by domain, show the number of answers and the prompts, and link the cited page.
- Filter by provider and date range; export the table as CSV.
- Mark domains already in an "ignore" list in .env (your own site, social networks).
- Update the README.

What you lose

Google AI Overviews and AI Mode tracking
Answers collected from the public chat interfaces in different countries
Sentiment, reputation and recommendation reports, and the GEO Writer drafts
AI-traffic and AI-crawler analytics and connectors for Looker Studio, Power BI and Tableau
A hosted history that keeps running without your machine and your API budget

Why people still pay

Marketing teams want a dashboard they can screenshot into a report, covering the Google AI surfaces and several countries, with someone else keeping the collection running as the apps change. The recommendations and drafting tools turn the numbers into a task list, which an API script does not.

Moatinfrastructure scale execution polish integrations

Do the math

LLM Pulse or your own copy?

Over

Your usage

Count my time12h ×$/h

LLM Pulse, 1 yr$660

Your own copy, 1 yr$280running + build session

You keep$380pays for itself in the first month

Estimate · 50 prompts, three runs each, three providers, weekly: about 1,900 calls a month. Search-enabled calls are billed per call plus tokens, so check each provider's current prices; it runs on your machine. The build is token usage on your agent plan.

Leaving LLM Pulse

Take your data, then cancel.

1 · Take your data out of LLM Pulse

1Open a data page such as Prompts, Citations or Sentiment, click Export table and choose CSV or Excel. Download it from the notification or from Exports in the profile menu; links expire, so download promptly.
2Open in Google Sheets from the same menu creates a spreadsheet in your Drive.
3For reports, click Download report (PDF) in the page header of Overview, Prompts, Recommendations, Sentiment and others; files stay available for 7 days.
4Everything else is reachable through the REST API, CLI or MCP with an API key from the profile menu.

2 · Cancel LLM Pulse

1Click your name at the bottom left of the sidebar, then Plans & Billing (owners and members with billing access).
2Click Manage Subscription to open the billing portal and cancel there; the page then shows Active and Renewal canceled.
3You keep your plan until the period ends; support can point you to the controls but can't cancel for you.
4To resume, use the same page to reactivate.

Your answer history is the valuable part. Export prompts, citations and sentiment tables before the period ends.
Paid yearly time isn't refunded when you switch from yearly to monthly. The billing help page describes no general refund policy: open a support ticket.
If you are over the project limit when you drop plans, only your oldest projects keep being tracked.
Trial: 14 days on weekly Starter, Growth and Scale only; daily plans and Scale+ start paid.

Checked 2026-10-19 against LLM Pulse's docs: 1, 2, 3, 4, 5

The escape hatch

Don't feel like building it? These are already free.

AnsvisorOpen sourceMIT platform that tracks prompts, mentions, citations and competitors across several AI assistants; self-host for free or use its cloud.★ 148 · active Oct 2026 · docker to self-host
WebVisit ↗GetCitoOpen sourceOpen-source AI visibility tracker with a hosted demo; the repository declares no licence file GitHub could detect, so check it before relying on it.★ 493 · active Sep 2026 · self-host, real ops
WebVisit ↗PotatoOpen sourceMIT command-line check that measures brand mentions and citations in Claude's web-search answers; runs locally and covers one engine.★ 166 · active Jun 2026 · one-click install
WebVisit ↗

All 4 free alternatives to LLM Pulse ›

The numbers

LLM Pulse pricing

Plan	Monthly	Annual, per month	What you get
Starter	$55	$45.84	50 tracked prompts, 1 project, 10 competitors per project, 3 GEO Writer tasks, unlimited team members, 5 AI models included.
Growth	$111.13	$92.61	150 tracked prompts, 2 projects, 15 competitors per project, 5 GEO Writer tasks; adds sentiment, owned media, AI traffic analytics and the LLM Pulse Agent.
Scale	$335.63	$279.69	450 tracked prompts, 5 projects, 20 competitors per project, 15 GEO Writer tasks; adds shopping and paid-ads tracking, reputation, GEO testing, custom reports, API, Looker Studio connector and CLI.
Scale+	$672.38	$560.31	1,200 tracked prompts and 10 projects.
Scale++	$1,345.88	$1,121.56	2,400 tracked prompts and 15 projects.
Enterprise	custom	—	Custom prompts and projects, white label, embedded dashboards, dedicated account manager, custom integrations, invoice billing.

Billing: Monthly or yearly; yearly costs ten times the monthly price (two months free). Prices are in euros excluding VAT and Stripe converts cards at its live rate, so the dollar figures here are ECB-rate estimates. The listed prices are for the default weekly tracking; daily tracking is priced separately and its prices were not read here. Upgrades apply at once with proration; downgrades wait until renewal. Cancel through Manage Subscription and keep access to the end of the period.
Hidden costs: Each extra tracked prompt costs €1 a month, an extra project €50 a month and an extra AI model from €8 a month (Claude, Copilot, Grok and others are paid add-ons). The five included models are ChatGPT, Perplexity, Google AI Mode, Google AI Overviews and Gemini.
Source: llmpulse.ai

Also on the list

Semrush$139/mo · SEO & marketingNOT REALLY

Ahrefs$29/mo · SEO & marketingNOT REALLY

AccuRanker$224/mo · SEO & marketingKINDA

Questions

LLM Pulse, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

No free plan. New customers may get a 14-day trial on weekly Starter, Growth or Scale; it needs a payment method.

Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: Google AI Overviews and AI Mode tracking, Answers collected from the public chat interfaces in different countries.

About $55 a month on the Starter (weekly tracking) plan, $660 a year (checked 2026-10-19).

Google AI Overviews and AI Mode tracking; Answers collected from the public chat interfaces in different countries; Sentiment, reputation and recommendation reports, and the GEO Writer drafts; AI-traffic and AI-crawler analytics and connectors for Looker Studio, Power BI and Tableau; A hosted history that keeps running without your machine and your API budget. If any of those matter to you, keep paying.